Prediction under uncertainty as a boundary problem: A general formulation using Iterative Closed Question Modelling

Introductory overview of identifiability analysis: A guide to evaluating whether you have the right type of data for your modeling purpose

“…As foreshadowed in Section 2.3.1, this is the approach preferred by Bayesian inference, which focuses on identifying a "credible region" of parameters -within which an unobserved parameter falls with a given (subjective) probability. It can also be approached more directly as a "set membership" problem, as notably used in the Generalised likelihood uncertainty estimation (GLUE) approach where only the set of "behavioural" parameters satisfying pre-defined constraints are retained from the prior set (Beven, 2006;Guillaume et al, 2015). It is worth repeating that even if the concept of a unique parameter vector is no longer used in parameter estimation, it is still useful in the context of identifiability analysis -in helping to rule out that non-uniqueness is due to sources I and II -which is often avoidable.…”

“…Consider whether non-identifiability can be tolerated: Uncertainty due to identifiability issues (whether explicitly quantified or not) can be assessed in terms of its risk, that is, its effect on the final product of the analysis, such as quantities of predictive interest for decision making. For example, if the uncertainty induced by non-identifiability does not change a decision, then perhaps it can be ignored (Guillaume et al, 2015). This is typically the default approach if modelers are aware of identifiability issues.…”

A bricolage-style exploratory scenario analysis to manage uncertainty in socio-environmental systems modeling: investigating integrated water management options

“…Another option is to look at what can be said about a problem despite uncertainties, i.e. what holds true in all the scenarios considered plausible (Fu & Guillaume, 2014;Guillaume et al, 2015), or what is "robust" by some metric (Giuliani & Castelletti, 2016;Herman et al, 2015;Kwakkel, Eker, et al, 2016;McPhail et al, 2018). An example of these approaches is robust decision making (Hall et al, 2012;Lempert et al, 2003), which iteratively searches for actions with robust outcomes and explores the circumstances leading to failure.…”